dma-buf: Check status of enable-signaling bit on debug
Subsystems (3)
cover dma-buf: Check status of enable-signaling bit on debug
Fence signaling must be enabled to make sure that the dma_fence_is_signaled() and dma_fence_is_signaled_locked() function ever returns true. Since drivers and implementations sometimes mess this up, this ensures correct behaviour when DEBUG_WW_MUTEX_SLOWPATH is used during debugging. This should make any implementation bugs resulting in not signaled fences much more obvious.
Arvind Yadav (3):
[PATCH 1/3] dma-buf: Remove the signaled bit status check [PATCH 2/3] dma-buf: Enable signaling on fence for sw_sync [PATCH 3/3] dma-buf: Check status of enable-signaling bit on debug
drivers/dma-buf/dma-fence.c | 5 ----- drivers/dma-buf/sw_sync.c | 2 ++ include/linux/dma-fence.h | 5 +++++ 3 files changed, 7 insertions(+), 5 deletions(-)
Signature · 1 line
2.25.1
PATCH 1/3 dma-buf: Remove the signaled bit status check
Remove the extra signaled bit status check because
it is returning early when the fence is already signaled and __dma_fence_enable_signaling is checking the status of signaled bit again.
--- drivers/dma-buf/dma-fence.c | 5 ----- 1 file changed, 5 deletions(-)
drivers/dma-buf/dma-fence.c +0 -5
@@ -648,11 +648,6 @@ int dma_fence_add_callback(struct dma_fence *fence, struct dma_fence_cb *cb,if (WARN_ON (!fence || !func))return -EINVAL ; -if (test_bit (DMA_FENCE_FLAG_SIGNALED_BIT , &fence->flags)) { -INIT_LIST_HEAD (&cb->node); -return -ENOENT ; - } -spin_lock_irqsave (fence->lock, flags);if (__dma_fence_enable_signaling (fence)) {
Signature · 1 line
2.25.1
Re: [PATCH 1/3] dma-buf: Remove the signaled bit status check
Am 27.09.22 um 19:24 schrieb Arvind Yadav:
Quoted 6 lines
Remove the extra signaled bit status check because
it is returning early when the fence is already signaled and__dma_fence_enable_signalingis checking the status of signaled bit again.Signed-off-by Arvind Yadav <[email protected]>
Quoted 20 lines
--- drivers/dma-buf/dma-fence.c | 5 ----- 1 file changed, 5 deletions(-)drivers/dma-buf/dma-fence.c +0
-5@@ -648,11 +648,6 @@ int dma_fence_add_callback(struct dma_fence *fence, struct dma_fence_cb *cb,if (WARN_ON (!fence || !func))return -EINVAL ; -if (test_bit (DMA_FENCE_FLAG_SIGNALED_BIT , &fence->flags)) { -INIT_LIST_HEAD (&cb->node); -return -ENOENT ; - } -spin_lock_irqsave (fence->lock, flags);if (__dma_fence_enable_signaling (fence)) {
PATCH 2/3 dma-buf: Enable signaling on fence for sw_sync
In reply to Arvind YadavHere's enabling software signaling on fence for sw_sync.
--- drivers/dma-buf/sw_sync.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/dma-buf/sw_sync.c +2 -0
@@ -244,6 +244,8 @@ static struct sync_pt *sync_pt_create(struct sync_timeline *obj, obj->context, value);INIT_LIST_HEAD (&pt->link); +dma_fence_enable_sw_signaling (&pt->base); +spin_lock_irq (&obj->lock);if (!dma_fence_is_signaled_locked (&pt->base)) {struct rb_node **p = &obj->pt_tree.rb_node;
Signature · 1 line
2.25.1
Re: [PATCH 2/3] dma-buf: Enable signaling on fence for sw_sync
Am 27.09.22 um 19:24 schrieb Arvind Yadav:
Quoted 3 lines
Here's enabling software signaling on fence for
sw_sync.Signed-off-by Arvind Yadav <[email protected]>
Quoted 17 lines
--- drivers/dma-buf/sw_sync.c | 2 ++ 1 file changed, 2 insertions(+)drivers/dma-buf/sw_sync.c +2
-0@@ -244,6 +244,8 @@ static struct sync_pt *sync_pt_create(struct sync_timeline *obj, obj->context, value);INIT_LIST_HEAD (&pt->link); +dma_fence_enable_sw_signaling (&pt->base); +spin_lock_irq (&obj->lock);if (!dma_fence_is_signaled_locked (&pt->base)) {struct rb_node **p = &obj->pt_tree.rb_node;
PATCH 3/3 dma-buf: Check status of enable-signaling bit on debug
In reply to Arvind YadavFence signaling must be enabled to make sure that
the dma_fence_is_signaled_locked() function ever returns true. Since drivers and implementations sometimes mess this up, this ensures correct behaviour when DMABUF_DEBUG_ENABLE_SIGNALING is used during debugging.
This should make any implementation bugs resulting in not signaled fences much more obvious.
--- include/linux/dma-fence.h | 5 +++++ 1 file changed, 5 insertions(+)
include/linux/dma-fence.h +5 -0
@@ -398,6 +398,11 @@ void dma_fence_enable_sw_signaling(struct dma_fence *fence);static inline bool dma_fence_is_signaled_locked (struct dma_fence *fence) { +#ifdef CONFIG_DMABUF_DEBUG_ENABLE_SIGNALING +if (!test_bit (DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT , &fence->flags)) +return false; +#endif +if (test_bit (DMA_FENCE_FLAG_SIGNALED_BIT , &fence->flags))return true;
Signature · 1 line
2.25.1
Christian König
maintainerAm 27.09.22 um 19:24 schrieb Arvind Yadav:
Quoted 7 lines
Fence signaling must be enabled to make sure that
thedma_fence_is_signaled_locked()function ever returns true. Since drivers and implementations sometimes mess this up, this ensures correct behaviour whenDMABUF_DEBUG_ENABLE_SIGNALINGis used during debugging.
This should make any implementation bugs resulting in not signaled fences much more obvious.
Are all IGT tests now passing with this? That would be a bit unfortunate because it means we still have missed the bug in drm_syncobj.
Christian.
Quoted 22 lines
Signed-off-by Arvind Yadav <[email protected]>--- include/linux/dma-fence.h | 5 +++++ 1 file changed, 5 insertions(+)include/linux/dma-fence.h +5
-0@@ -398,6 +398,11 @@ void dma_fence_enable_sw_signaling(struct dma_fence *fence);static inline bool dma_fence_is_signaled_locked (struct dma_fence *fence) { +#ifdef CONFIG_DMABUF_DEBUG_ENABLE_SIGNALING +if (!test_bit (DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT , &fence->flags)) +return false; +#endif +if (test_bit (DMA_FENCE_FLAG_SIGNALED_BIT , &fence->flags))return true;
Quoted 11 lines · Christian König
Am 27.09.22 um 19:24 schrieb Arvind Yadav:
Quoted 7 lines
Fence signaling must be enabled to make sure that
thedma_fence_is_signaled_locked()function ever returns true. Since drivers and implementations sometimes mess this up, this ensures correct behaviour whenDMABUF_DEBUG_ENABLE_SIGNALINGis used during debugging.
This should make any implementation bugs resulting in not signaled fences much more obvious.Are all IGT tests now passing with this? That would be a bit unfortunate because it means we still have missed the bug in
drm_syncobj.
IGT has these test cases related to syncobj (syncobj_basic, syncobj_timeline and syncobj_wait)and all are passing.
I will check syncobj and let you know.
~Arvind
Quoted 25 lines
Christian.
Quoted 22 lines
Signed-off-by Arvind Yadav <[email protected]>---
include/linux/dma-fence.h| 5 +++++1 file changed, 5 insertions(+)include/linux/dma-fence.h +0
-0@@ -398,6 +398,11 @@ void dma_fence_enable_sw_signaling(struct dma_fence *fence);static inline bool
dma_fence_is_signaled_locked(struct dma_fence *fence){include/linux/dma-fence.h +5
-0+#ifdef CONFIG_DMABUF_DEBUG_ENABLE_SIGNALING + if (!test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &fence->flags)) + return false; +#endif +if (
test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) return true;
Christian König
maintainerAm 29.09.22 um 20:30 schrieb Yadav, Arvind:
Quoted 11 lines · Christian König
Quoted 9 lines
Am 27.09.22 um 19:24 schrieb Arvind Yadav:
Quoted 7 lines
Fence signaling must be enabled to make sure that
thedma_fence_is_signaled_locked()function ever returns true. Since drivers and implementations sometimes mess this up, this ensures correct behaviour whenDMABUF_DEBUG_ENABLE_SIGNALINGis used during debugging.
This should make any implementation bugs resulting in not signaled fences much more obvious.
Quoted 5 lines
Quoted 2 lines
Are all IGT tests now passing with this? That would be a bit unfortunate because it means we still have missed the bug in
drm_syncobj.IGT has these test cases related to syncobj (
syncobj_basic,syncobj_timelineand syncobj_wait)and all are passing.I will check syncobj and let you know.
Maybe CC the Intel list and let their CI systems take a look. That's usually rather valuable.
Thanks,
Christian.
Quoted 28 lines
~Arvind
Quoted 25 lines
Christian.
Quoted 22 lines
Signed-off-by Arvind Yadav <[email protected]>---
include/linux/dma-fence.h| 5 +++++1 file changed, 5 insertions(+)include/linux/dma-fence.h +0
-0@@ -398,6 +398,11 @@ void dma_fence_enable_sw_signaling(struct dma_fence *fence);static inline bool
dma_fence_is_signaled_locked(struct dma_fence *fence){include/linux/dma-fence.h +5
-0+#ifdef CONFIG_DMABUF_DEBUG_ENABLE_SIGNALING + if (!test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &fence->flags)) + return false; +#endif +if (
test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) return true;
Quoted 12 lines · Christian König
Am 29.09.22 um 20:30 schrieb Yadav, Arvind:
Quoted 11 lines · Christian König
Quoted 9 lines
Am 27.09.22 um 19:24 schrieb Arvind Yadav:
Quoted 7 lines
Fence signaling must be enabled to make sure that
thedma_fence_is_signaled_locked()function ever returns true. Since drivers and implementations sometimes mess this up, this ensures correct behaviour whenDMABUF_DEBUG_ENABLE_SIGNALINGis used during debugging.
This should make any implementation bugs resulting in not signaled fences much more obvious.
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines
Are all IGT tests now passing with this? That would be a bit unfortunate because it means we still have missed the bug in
drm_syncobj.IGT has these test cases related to syncobj (
syncobj_basic,syncobj_timelineand syncobj_wait)and all are passing.I will check syncobj and let you know.
Maybe CC the Intel list and let their CI systems take a look. That's usually rather valuable.
There is one IGT subtest is failing which is related to syncobj. I will fix this and submit the patch.
igt_subtest("host-signal-points")
test_host_signal_points(fd);
Quoted 33 lines
Thanks,
Christian.Quoted 28 lines
~Arvind
Quoted 25 lines
Christian.
Quoted 22 lines
Signed-off-by Arvind Yadav <[email protected]>---
include/linux/dma-fence.h| 5 +++++1 file changed, 5 insertions(+)include/linux/dma-fence.h +0
-0@@ -398,6 +398,11 @@ void dma_fence_enable_sw_signaling(struct dma_fence *fence);static inline bool
dma_fence_is_signaled_locked(struct dma_fence *fence){include/linux/dma-fence.h +5
-0+#ifdef CONFIG_DMABUF_DEBUG_ENABLE_SIGNALING + if (!test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &fence->flags)) + return false; +#endif +if (
test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) return true;